Commodity recommendation method, device, electronic device and medium based on artificial intelligence

By calculating the time impact factor and interest preference factor, updating user data, and recommending products based on retention prediction models, the problem of low prediction accuracy of Logistic regression model in the prior art is solved, and the effect of product recommendation is improved.

CN114399368BActive Publication Date: 2024-11-22PING AN TECH (SHENZHEN) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210081099.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-01-24
Publication Date
2024-11-22
Estimated Expiration
2042-01-24

AI Technical Summary

Technical Problem

In the prior art, the accuracy of user retention prediction through the Logistic regression model is low, resulting in poor product recommendation results.

Method used

By obtaining the initial user data of the target user group, calculating the time impact factor and interest preference factor, updating the user data, making retention prediction based on the retention prediction model, and product recommendations are made based on the similarity.

Benefits of technology

It improves the accuracy and effectiveness of product recommendations, and improves the accuracy of similarity and enhances the accuracy of recommendations by updating user data and interest preference calculations in real time.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114399368B_ABST
    Figure CN114399368B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of artificial intelligence technology, and provides a commodity recommendation method, device, electronic device and medium based on artificial intelligence. After obtaining the initial user data of the first and second target user groups, the initial user data of the first and second target user groups are updated to the target user data by calculating the time influence factor, thereby correcting the problem that the target user data cannot be updated in real time due to the time factor. Further, by calculating the interest preference factors of the first target user data and the second target user data, the similarity of the two corresponding target user data is calculated based on the interest preference factors, and the accuracy of the similarity is high. The retention prediction model trained based on the second target user data is used to perform retention prediction on the first target user data to obtain a retention prediction label. Finally, the retention prediction label and the similarity are combined to recommend commodities to the first target user group, and the commodity recommendation effect is better.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based commodity recommendation method, device, electronic device and medium. Background Art

[0002] User retention is a common big data mining application scenario. Users who start using an application within a certain period of time and continue to use the application after a period of time are considered retained users. Retained users reflect the quality of the product and the ability to retain users.

[0003] In the process of realizing the present invention, the inventors found that in the prior art, a Logistic regression model is trained by obtaining data of different attributes of users to perform retention prediction, and products are recommended based on the probability of retention prediction. However, the regression coefficient of the Logistic regression model is an estimated value obtained by maximum likelihood estimation using existing user data, and the Logistic regression model is optimized based on the estimated value. Therefore, when the trained Logistic regression model is used to perform retention prediction for target users, the accuracy of the retention prediction is low, resulting in poor recommendation effect. Summary of the invention

[0004] In view of the above content, it is necessary to propose a product recommendation method, device, electronic device and medium based on artificial intelligence, which can improve the effect of product recommendation.

[0005] A first aspect of the present invention provides a commodity recommendation method based on artificial intelligence, the method comprising:

[0006] Acquire first initial user data of a first target user group, and acquire second initial user data of a second target user group;

[0007] Calculate a time impact factor and use the time impact factor to update the first initial user data to first target user data and update the second initial user data to second target user data;

[0008] Calculating interest preference factors of the first target user data and the second target user data, and calculating the similarity between the first target user data and the second target user data based on the interest preference factors;

[0009] Training a retention prediction model based on the second target user data, and using the retention prediction model to perform retention prediction on the first target user data to obtain a retention prediction label;

[0010] Recommend products to the first target user group based on the retention prediction label and the similarity.

[0011] According to an optional implementation manner of the present invention, calculating the time impact factor and using the time impact factor to update the first initial user data to the first target user data and updating the second initial user data to the second target user data includes:

[0012] Obtaining a first registration time of the first target user and obtaining a second registration time of the second target user;

[0013] Calculating a first time impact factor of the first initial user data corresponding to the first registration time, and calculating a second time impact factor of the second initial user data corresponding to the second registration time;

[0014] The first time impact factor is used to update the first initial user data to the first target user data, and the second time impact factor is used to update the second initial user data to the second target user data.

[0015] According to an optional implementation manner of the present invention, the calculating the interest preference factors of the first target user data and the second target user data, and calculating the similarity between the corresponding first target user data and the second target user data based on the interest preference factors includes:

[0016] Calculating a first number of identical target user data in the first target user data and the second target user data;

[0017] Calculating a second amount of total target user data of the first target user data and the second target user data;

[0018] Calculate an interest preference factor based on the first number and the second number;

[0019] Calculating the Pearson correlation coefficient between the first target user data and the second target user data;

[0020] The interest preference factor is used to correct the corresponding Pearson correlation coefficient to obtain the similarity.

[0021] According to an optional implementation manner of the present invention, calculating the Pearson correlation coefficient between the first target user data and the second target user data includes:

[0022] Calculating first mean data according to the first target user data, and calculating second mean data according to the second target user data;

[0023] Calculating first difference data according to the first target user data and the first mean data, and calculating second difference data according to the second target user data and the second mean data;

[0024] The Pearson correlation coefficient between the first target user data and the second target user data is calculated based on the first difference data and the second difference data.

[0025] According to an optional implementation manner of the present invention, the acquiring first initial user data of the first target user group and the acquiring second initial user data of the second target user group include:

[0026] In response to the received product recommendation instruction, parsing the method body of the product recommendation instruction to obtain information carried by the product recommendation instruction;

[0027] Acquire a preset first tag of the first target user group and a preset second tag of the second target user group from the information;

[0028] Constructing a first regular expression according to the preset first tag and constructing a second regular expression according to the preset second tag;

[0029] Determine a first database identifier by using the first regular expression, and determine data traversed from the first database corresponding to the first database identifier as first initial user data;

[0030] The second regular expression is used to determine a second database identifier, and the data traversed from the second database corresponding to the second database identifier is determined as second initial user data.

[0031] According to an optional implementation manner of the present invention, the training of the retention prediction model based on the second target user data includes:

[0032] Acquire third initial user data of a third target user group, and obtain third target user data according to the third initial user data;

[0033] Using the second target user data and the third target user data as input data of the model, and using the retained labels corresponding to the second target user data and the third target user data as training targets of the model;

[0034] According to the input data of the model and the training objectives of the model, training is performed based on the gradient boosting decision tree algorithm to obtain the retention prediction model.

[0035] According to an optional implementation of the present invention, the recommending products for the first target user group according to the retention prediction label and the similarity includes:

[0036] Obtaining a target retention prediction label from the retention prediction labels;

[0037] Acquire a first target user corresponding to the target retention prediction label in the first target user group as a user to be recommended;

[0038] Sorting the similarities corresponding to the users to be recommended;

[0039] Determine the second target users corresponding to the top N similarities as matching users of the user to be recommended;

[0040] Recommend products to the user to be recommended based on the products purchased by the matching user.

[0041] A second aspect of the present invention provides a commodity recommendation device based on artificial intelligence, the device comprising:

[0042] An acquisition module, used to acquire first initial user data of a first target user group, and to acquire second initial user data of a second target user group;

[0043] An updating module, configured to calculate a time impact factor and use the time impact factor to update the first initial user data to first target user data and to update the second initial user data to second target user data;

[0044] a calculation module, configured to calculate an interest preference factor of the first target user data and the second target user data, and calculate a similarity between the first target user data and the second target user data based on the interest preference factor;

[0045] A prediction module, configured to train a retention prediction model based on the second target user data, and use the retention prediction model to perform retention prediction on the first target user data to obtain a retention prediction label;

[0046] A recommendation module is used to recommend products to the first target user group based on the retention prediction label and the similarity.

[0047] A third aspect of the present invention provides an electronic device, comprising a processor, wherein the processor is configured to implement the artificial intelligence-based commodity recommendation method when executing a computer program stored in a memory.

[0048] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, and the computer program, when executed by a processor, implements the artificial intelligence-based commodity recommendation method.

[0049] In summary, the artificial intelligence-based commodity recommendation method, device, electronic device and medium of the present invention, after obtaining the first initial user data of the first target user group and the second initial user data of the second target user group, because the user's personal interests will change with the passage of time, the first initial user data is updated to the first target user data and the second initial user data is updated to the second target user data by calculating the time influence factor, thereby correcting the problem that the target user data cannot be updated in real time due to the time factor, and further by calculating the interest preference factors of the first target user data and the second target user data, the potential interests and hobbies between the target users corresponding to the two target user data are effectively characterized, and the similarity of the corresponding two target user data is calculated based on the interest preference factors, and the accuracy of the similarity is high. The retention prediction model trained based on the second target user data is used to perform retention prediction on the first target user data to obtain a retention prediction label. Finally, by combining the retention prediction label and the similarity factors, commodity recommendations are made for the first target user group, and the effect of commodity recommendation is better. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a flowchart of an artificial intelligence-based product recommendation method provided in Example 1 of the present invention.

[0051] Figure 2 It is a structural diagram of an artificial intelligence-based commodity recommendation device provided in Embodiment 2 of the present invention.

[0052] Figure 3 It is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0053] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0054] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing embodiments in an optional embodiment and are not intended to limit the present invention.

[0055] The artificial intelligence-based commodity recommendation method provided in the embodiment of the present invention is executed by an electronic device, and accordingly, the artificial intelligence-based commodity recommendation device runs in the electronic device.

[0056] Embodiment 1

[0057] Figure 1 The flowchart of the product recommendation method based on artificial intelligence provided by the first embodiment of the present invention is shown in FIG. The product recommendation method based on artificial intelligence specifically includes the following steps. According to different requirements, the order of the steps in the flowchart can be changed, and some steps can be omitted.

[0058] S11, obtaining first initial user data of a first target user group, and obtaining second initial user data of a second target user group.

[0059] The embodiment of the present application recommends products to the first target user group based on the second target user group, where the products may include insurance products, learning courses, audio and video, blogs, etc. The embodiment of the present application does not impose any specific restrictions on the types of products.

[0060] To facilitate understanding of the embodiments of the present application, the following takes insurance product recommendation as an example, wherein the first target user group refers to historical users who are not currently insured, and the second target user group refers to historical users who have currently renewed their insurance. That is, the embodiments of the present application do not perform retention predictions on all historical users and recommend insurance products based on the retention prediction probability.

[0061] In the application scenario of insurance product recommendation, the initial user data may include, but is not limited to: historical insurance product purchase information, historical insurance agent information, online activity information, etc.

[0062] In an optional implementation, the first initial user data and the second initial user data may include: the time interval of user activities, the time difference between the time of the user's last activity and the current time, the user's registration time, etc.

[0063] For example, the time interval between users launching insurance apps, the time interval between interactions and likes, the time interval between browsing product details, etc. The time difference between the last time a user launched an insurance app and the current time, the time difference between the last time a user interacted and liked and the current time, and the time difference between the time a user browsed product details and the current time.

[0064] Intuitively, the time interval between user activities is the best feature to describe the user's activity level. The longer the time interval between activities, the less active the user is and the greater the possibility of churn. On the contrary, the shorter the time interval between activities, the more active the user is and the less likely the user is to churn. Considering that user activities have a certain stickiness, the time difference between the last time the user was active and the current time can be used to describe the stickiness of user activities. The larger the time difference, the smaller the stickiness and the less likely the user is to be active again. On the contrary, the smaller the time difference, the more likely the user is to be active again.

[0065] In an optional implementation, the acquiring of first initial user data of the first target user group and the acquiring of second initial user data of the second target user group include:

[0066] In response to the received product recommendation instruction, parsing the method body of the product recommendation instruction to obtain information carried by the product recommendation instruction;

[0067] Acquire a preset first tag of the first target user group and a preset second tag of the second target user group from the information;

[0068] Constructing a first regular expression according to the preset first tag and constructing a second regular expression according to the preset second tag;

[0069] Determine a first database identifier by using the first regular expression, and determine data traversed from the first database corresponding to the first database identifier as first initial user data;

[0070] The second regular expression is used to determine a second database identifier, and the data traversed from the second database corresponding to the second database identifier is determined as second initial user data.

[0071] The commodity recommendation instruction may be triggered by a department head who needs to make a commodity recommendation, and the present invention is not limited thereto.

[0072] The product recommendation instruction is essentially a code. According to the code writing principle, the content between {} is called the method body.

[0073] Different target user groups correspond to different preset tags. The first target user group corresponds to a preset first tag, and the second target user group corresponds to a preset second tag. The preset first tag and the preset second tag can be custom configured.

[0074] The preset tag and the database identifier have a one-to-one correspondence. For example, the preset tag may be ID, and a regular expression ID() is further established with the preset tag, and traversal is performed with ID().

[0075] The above optional implementation manner constructs a regular expression based on a preset tag, uses the regular expression to quickly determine a database identifier, and further uses the database identifier to call a corresponding database, so as to quickly and accurately obtain the initial user data of the target user group from the corresponding database.

[0076] S12, calculating a time impact factor and using the time impact factor to update the first initial user data into first target user data and to update the second initial user data into second target user data.

[0077] Users' personal interests will continue to change with the passage of time. For example, the demand for purchasing insurance products and the types of insurance products will continue to change with the passage of time. By calculating the time impact factor, the initial user data is adjusted to the target user data, and the similarity is calculated based on the target user data.

[0078] In an optional implementation, calculating the time impact factor and using the time impact factor to update the first initial user data to the first target user data and updating the second initial user data to the second target user data includes:

[0079] Obtaining a first registration time of the first target user and obtaining a second registration time of the second target user;

[0080] Calculating a first time impact factor of the first initial user data corresponding to the first registration time, and calculating a second time impact factor of the second initial user data corresponding to the second registration time;

[0081] The first time impact factor is used to update the first initial user data to the first target user data, and the second time impact factor is used to update the second initial user data to the second target user data.

[0082] Since the first target user group and the second target user group are users who have purchased insurance products in the past, each first target user and each second target user will log in to the insurance product application to register when purchasing an insurance product.

[0083] The time difference between the registration time and the current time is calculated, and the time difference is used as the time impact factor. Based on the product of the time impact factor and the initial user data, the target user data is obtained, thereby achieving the purpose of updating the initial user data to the target user data.

[0084] In the above optional implementation, the time impact factor is calculated by the time difference between the target user's registration time and the current time, and the corresponding initial user data is updated according to the time impact factor, so that the obtained target user data effectively integrates the time factor.

[0085] S13, calculating the interest preference factors of the first target user data and the second target user data, and calculating the similarity between the corresponding first target user data and the second target user data based on the interest preference factors.

[0086] Generally speaking, the common user data between different target user groups is relatively sparse, but the values ​​of the user data may be relatively close, resulting in a relatively close similarity in the final calculation. This situation may overestimate the similarity of users. For example, if two users are interested in a small number of insurance products at the same time, their interests and hobbies may be very dissimilar. Therefore, when using the Pearson correlation coefficient to calculate the similarity of target users between two different target user groups, it is necessary to consider the common interests and hobbies of the target users to determine the interest preference factor.

[0087] The greater the similarity between the first target user data and the second target user data, the more consistent the preferences of the corresponding first target user and the second target user are; the smaller the similarity between the first target user data and the second target user data, the less consistent the preferences of the corresponding first target user and the second target user are.

[0088] In an optional implementation, the calculating the interest preference factors of the first target user data and the second target user data, and calculating the similarity between the corresponding first target user data and the second target user data based on the interest preference factors includes:

[0089] Calculating a first number of identical target user data in the first target user data and the second target user data;

[0090] Calculating a second amount of total target user data of the first target user data and the second target user data;

[0091] Calculate an interest preference factor based on the first number and the second number;

[0092] Calculating the Pearson correlation coefficient between the first target user data and the second target user data;

[0093] The interest preference factor is used to correct the corresponding Pearson correlation coefficient to obtain the similarity.

[0094] Exemplarily, assuming that the first target user data D1 includes 10 data fields, the second target user data D2 includes 10 data fields, and 5 data fields in the first target user data D1 and the second target user data D2 are the same, the interest preference factor is 5 / 20=0.25.

[0095] After calculating the Pearson correlation coefficient between the first target user data and the second target user data, the Pearson correlation coefficient and the corresponding interest preference factor are weighted, and the obtained weighted value is used as the similarity between the first target user data and the second target user data.

[0096] The above optional implementation method can effectively characterize the potential interests and hobbies between the target users corresponding to the two target user data by calculating the number of identical target user data in the two target user data, so that when the Pearson correlation coefficient is calculated, the Pearson correlation coefficient is corrected to obtain a similarity with higher accuracy, which helps to improve the effect of subsequent product recommendations.

[0097] In an optional implementation, calculating the Pearson correlation coefficient between the first target user data and the second target user data includes:

[0098] Calculating first mean data according to the first target user data, and calculating second mean data according to the second target user data;

[0099] Calculating first difference data according to the first target user data and the first mean data, and calculating second difference data according to the second target user data and the second mean data;

[0100] The Pearson correlation coefficient between the first target user data and the second target user data is calculated based on the first difference data and the second difference data.

[0101] It should be noted that before calculating the first mean data and the second mean data, the first target user data and the second target user data need to be unified in terms of dimensions, that is, non-numeric data in the first target user data and the second target user data need to be mapped to numeric data.

[0102] Each data field in the first target user data is subtracted from the first mean data to obtain first difference data, and each data field in the second target user data is subtracted from the second mean data to obtain second difference data.

[0103] In the above optional implementation, by calculating the mean data of each target user data, then using the target user data to subtract the corresponding mean data, and finally calculating the Pearson correlation coefficient based on the obtained difference data, not only the directionality between the data is considered, but also the specific size of the data is taken into account. Therefore, the calculated Pearson correlation coefficient is more accurate, thereby making the calculated similarity more accurate.

[0104] S14: training a retention prediction model based on the second target user data, and using the retention prediction model to perform retention prediction on the first target user data to obtain a retention prediction label.

[0105] In this embodiment, a logistic regression algorithm may be used to train the retention prediction model based on the second target user data.

[0106] The second target user data can be processed using a big data computing framework (mapreduce+hive). Specifically, hive can be used to collect the second target user data, the model can be encapsulated through mapreduce, and then hive can be used to verify or test the results predicted by the model.

[0107] In an optional implementation, the training of the retention prediction model based on the second target user data includes:

[0108] Acquire third initial user data of a third target user group, and obtain third target user data according to the third initial user data;

[0109] Using the second target user data and the third target user data as input data of the model, and using the retained labels corresponding to the second target user data and the third target user data as training targets of the model;

[0110] According to the input data of the model and the training target of the model, training is performed based on the Gradient Boosting Decision Tree (GBDT) algorithm to obtain the retention prediction model.

[0111] The third target user group refers to historical users who have given up on renewal. The third initial user data is the same as the first initial user data and the second initial user data, and may also include the time interval of user activities, the time difference between the time of the user's last activity and the current time, the user's registration time, etc. The third registration time of each of the third target users is obtained, and the third time impact factor of the corresponding third initial user data is calculated according to the third registration time, and the third time impact factor is used to update the third initial user data to the third target user data.

[0112] The retention label of the second target user data is recorded as "1", and the retention label of the third target user data is recorded as "0". The second target user data and the third target user data are used as training data sets to train the gradient boosting decision tree to obtain a retention prediction model, and then the retention prediction model is used to perform a binary classification prediction on the first target user data. That is, it is predicted whether the retention label of the first target user data is "1" or "0".

[0113] In the above optional implementation, the retention label is continuously fitted to the target for training. At the same time, the GBDT algorithm focuses on training samples with large residuals in each iteration of the iteration process, and then obtains multiple tree models. The final result is obtained by fusing multiple models, which fully considers the influence of residuals and can make the training effect of the retention prediction model better. When the retention prediction model with better training effect is used to perform retention prediction on the first target user data, a more accurate retention prediction label can be obtained.

[0114] S15: Recommend products to the first target user group based on the retention prediction label and the similarity.

[0115] Users' consumption behavior is an important channel for companies to make profits. The larger the user base, the greater the profit space for the company. When the rate of user loss is lower than the rate of acquiring new users, the number of users is increasing. Predicting possible retained users in advance and making product recommendations to retained users in a timely manner are of great practical significance for reducing user churn rate.

[0116] In an optional implementation, the recommending products to the first target user group according to the retention prediction tag and the similarity includes:

[0117] Obtaining a target retention prediction label from the retention prediction labels;

[0118] Acquire a first target user corresponding to the target retention prediction label in the first target user group as a user to be recommended;

[0119] Sorting the similarities corresponding to the users to be recommended;

[0120] Determine the second target users corresponding to the top N similarities as matching users of the user to be recommended;

[0121] Recommend products to the user to be recommended based on the products purchased by the matching user.

[0122] Among them, the target retention prediction label is "1".

[0123] Since the first target users with the target retention prediction label have a greater probability of being retained and becoming the main force of economic support for the enterprise, the factor only makes personalized recommendations of products to the first target users with the target retention prediction label in the first target user group.

[0124] The greater the similarity between the first target user group and the second target user group, the more similar their potential interests and hobbies are. Therefore, generating a product list based on the products purchased by the second target user group and recommending it to the first target user group can better meet the purchasing needs of the first target user group and achieve higher recommendation accuracy.

[0125] In the above optional implementation, by combining the retention prediction label and the similarity factor, product recommendations are made for the first target user group, and the effect of product recommendation is better.

[0126] In the embodiment of the present application, after obtaining the first initial user data of the first target user group and the second initial user data of the second target user group, since the personal interests of the users will change continuously with the passage of time, the first initial user data is updated to the first target user data and the second initial user data is updated to the second target user data by calculating the time influence factor, thereby correcting the problem that the target user data cannot be updated in real time due to the time factor, and further by calculating the interest preference factors of the first target user data and the second target user data, the potential interests and hobbies between the target users corresponding to the two target user data are effectively characterized, and the similarity of the corresponding two target user data is calculated based on the interest preference factor, and the accuracy of the similarity is high. The retention prediction model trained based on the second target user data is used to perform retention prediction on the first target user data to obtain a retention prediction label, and finally, by combining the retention prediction label and the similarity factors, product recommendations are performed for the first target user group, and the effect of product recommendation is better.

[0127] Embodiment 2

[0128] Figure 2 It is a structural diagram of an artificial intelligence-based commodity recommendation device provided in Embodiment 2 of the present invention.

[0129] In some embodiments, the artificial intelligence-based commodity recommendation device 20 may include a plurality of functional modules composed of computer program segments. The computer programs of the various program segments in the artificial intelligence-based commodity recommendation device 20 may be stored in a memory of an electronic device and executed by at least one processor to perform (see Figure 1 Description) Product recommendation function based on artificial intelligence.

[0130] In this embodiment, the artificial intelligence-based commodity recommendation device 20 can be divided into multiple functional modules according to the functions it performs. The functional modules may include: an acquisition module 201, an update module 202, a calculation module 203, a prediction module 204 and a recommendation module 205. The module referred to in the present invention refers to a series of computer program segments that can be executed by at least one processor and can perform fixed functions, which are stored in a memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0131] The acquisition module 201 is used to acquire first initial user data of a first target user group and to acquire second initial user data of a second target user group.

[0132] The embodiment of the present application recommends products to the first target user group based on the second target user group, where the products may include insurance products, learning courses, audio and video, blogs, etc. The embodiment of the present application does not impose any specific restrictions on the types of products.

[0133] To facilitate understanding of the embodiments of the present application, the following takes insurance product recommendation as an example, wherein the first target user group refers to historical users who are not currently insured, and the second target user group refers to historical users who have currently renewed their insurance. That is, the embodiments of the present application do not perform retention predictions on all historical users and recommend insurance products based on the retention prediction probability.

[0134] In the application scenario of insurance product recommendation, the initial user data may include, but is not limited to: historical insurance product purchase information, historical insurance agent information, online activity information, etc.

[0135] In an optional implementation, the first initial user data and the second initial user data may include: the time interval of user activities, the time difference between the time of the user's last activity and the current time, the user's registration time, etc.

[0136] For example, the time interval between users launching insurance apps, the time interval between interactions and likes, the time interval between browsing product details, etc. The time difference between the last time a user launched an insurance app and the current time, the time difference between the last time a user interacted and liked and the current time, and the time difference between the time a user browsed product details and the current time.

[0137] Intuitively, the time interval between user activities is the best feature to describe the user's activity level. The longer the time interval between activities, the less active the user is and the greater the possibility of churn. On the contrary, the shorter the time interval between activities, the more active the user is and the less likely the user is to churn. Considering that user activities have a certain stickiness, the time difference between the last time the user was active and the current time can be used to describe the stickiness of user activities. The larger the time difference, the smaller the stickiness and the less likely the user is to be active again. On the contrary, the smaller the time difference, the more likely the user is to be active again.

[0138] In an optional implementation, the acquiring module 201 acquires the first initial user data of the first target user group and acquires the second initial user data of the second target user group, including:

[0139] In response to the received product recommendation instruction, parsing the method body of the product recommendation instruction to obtain information carried by the product recommendation instruction;

[0140] Acquire a preset first tag of the first target user group and a preset second tag of the second target user group from the information;

[0141] Constructing a first regular expression according to the preset first tag and constructing a second regular expression according to the preset second tag;

[0142] Determine a first database identifier by using the first regular expression, and determine data traversed from the first database corresponding to the first database identifier as first initial user data;

[0143] The second regular expression is used to determine a second database identifier, and the data traversed from the second database corresponding to the second database identifier is determined as second initial user data.

[0144] The commodity recommendation instruction may be triggered by a department head who needs to make a commodity recommendation, and the present invention is not limited thereto.

[0145] The product recommendation instruction is essentially a code. According to the code writing principle, the content between {} is called the method body.

[0146] Different target user groups correspond to different preset tags. The first target user group corresponds to a preset first tag, and the second target user group corresponds to a preset second tag. The preset first tag and the preset second tag can be custom configured.

[0147] The preset tag and the database identifier have a one-to-one correspondence. For example, the preset tag may be ID, and a regular expression ID() is further established with the preset tag, and traversal is performed with ID().

[0148] The above optional implementation manner constructs a regular expression based on a preset tag, uses the regular expression to quickly determine a database identifier, and further uses the database identifier to call a corresponding database, so as to quickly and accurately obtain the initial user data of the target user group from the corresponding database.

[0149] The updating module 202 is configured to calculate a time impact factor and use the time impact factor to update the first initial user data to first target user data and to update the second initial user data to second target user data.

[0150] Users' personal interests will continue to change with the passage of time. For example, the demand for purchasing insurance products and the types of insurance products will continue to change with the passage of time. By calculating the time impact factor, the initial user data is adjusted to the target user data, and the similarity is calculated based on the target user data.

[0151] In an optional implementation, the updating module 202 calculates the time impact factor and uses the time impact factor to update the first initial user data to the first target user data and to update the second initial user data to the second target user data, including:

[0152] Obtaining a first registration time of the first target user and obtaining a second registration time of the second target user;

[0153] Calculating a first time impact factor of the first initial user data corresponding to the first registration time, and calculating a second time impact factor of the second initial user data corresponding to the second registration time;

[0154] The first time impact factor is used to update the first initial user data to the first target user data, and the second time impact factor is used to update the second initial user data to the second target user data.

[0155] Since the first target user group and the second target user group are users who have purchased insurance products in the past, each first target user and each second target user will log in to the insurance product application to register when purchasing an insurance product.

[0156] The time difference between the registration time and the current time is calculated, and the time difference is used as the time impact factor. Based on the product of the time impact factor and the initial user data, the target user data is obtained, thereby achieving the purpose of updating the initial user data to the target user data.

[0157] In the above optional implementation, the time impact factor is calculated by the time difference between the target user's registration time and the current time, and the corresponding initial user data is updated according to the time impact factor, so that the obtained target user data effectively integrates the time factor.

[0158] The calculation module 203 is used to calculate the interest preference factors of the first target user data and the second target user data, and calculate the similarity between the corresponding first target user data and the second target user data based on the interest preference factors.

[0159] Generally speaking, the common user data between different target user groups is relatively sparse, but the values ​​of the user data may be relatively close, resulting in a relatively close similarity in the final calculation. This situation may overestimate the similarity of users. For example, if two users are interested in a small number of insurance products at the same time, their interests and hobbies may be very dissimilar. Therefore, when using the Pearson correlation coefficient to calculate the similarity of target users between two different target user groups, it is necessary to consider the common interests and hobbies of the target users to determine the interest preference factor.

[0160] The greater the similarity between the first target user data and the second target user data, the more consistent the preferences of the corresponding first target user and the second target user are; the smaller the similarity between the first target user data and the second target user data, the less consistent the preferences of the corresponding first target user and the second target user are.

[0161] In an optional implementation, the calculation module 203 calculates the interest preference factors of the first target user data and the second target user data, and calculates the similarity between the first target user data and the second target user data based on the interest preference factors, including:

[0162] Calculating a first number of identical target user data in the first target user data and the second target user data;

[0163] Calculating a second amount of total target user data of the first target user data and the second target user data;

[0164] Calculate an interest preference factor based on the first number and the second number;

[0165] Calculating the Pearson correlation coefficient between the first target user data and the second target user data;

[0166] The interest preference factor is used to correct the corresponding Pearson correlation coefficient to obtain the similarity.

[0167] Exemplarily, assuming that the first target user data D1 includes 10 data fields, the second target user data D2 includes 10 data fields, and 5 data fields in the first target user data D1 and the second target user data D2 are the same, the interest preference factor is 5 / 20=0.25.

[0168] After calculating the Pearson correlation coefficient between the first target user data and the second target user data, the Pearson correlation coefficient and the corresponding interest preference factor are weighted, and the obtained weighted value is used as the similarity between the first target user data and the second target user data.

[0169] The above optional implementation method can effectively characterize the potential interests and hobbies between the target users corresponding to the two target user data by calculating the number of identical target user data in the two target user data, so that when the Pearson correlation coefficient is calculated, the Pearson correlation coefficient is corrected to obtain a similarity with higher accuracy, which helps to improve the effect of subsequent product recommendations.

[0170] In an optional implementation, calculating the Pearson correlation coefficient between the first target user data and the second target user data includes:

[0171] Calculating first mean data according to the first target user data, and calculating second mean data according to the second target user data;

[0172] Calculating first difference data according to the first target user data and the first mean data, and calculating second difference data according to the second target user data and the second mean data;

[0173] The Pearson correlation coefficient between the first target user data and the second target user data is calculated based on the first difference data and the second difference data.

[0174] It should be noted that before calculating the first mean data and the second mean data, the first target user data and the second target user data need to be unified in terms of dimensions, that is, non-numeric data in the first target user data and the second target user data need to be mapped to numeric data.

[0175] Each data field in the first target user data is subtracted from the first mean data to obtain first difference data, and each data field in the second target user data is subtracted from the second mean data to obtain second difference data.

[0176] In the above optional implementation, by calculating the mean data of each target user data, then using the target user data to subtract the corresponding mean data, and finally calculating the Pearson correlation coefficient based on the obtained difference data, not only the directionality between the data is considered, but also the specific size of the data is taken into account. Therefore, the calculated Pearson correlation coefficient is more accurate, thereby making the calculated similarity more accurate.

[0177] The prediction module 204 is used to train a retention prediction model based on the second target user data, and use the retention prediction model to perform retention prediction on the first target user data to obtain a retention prediction label.

[0178] In this embodiment, a logistic regression algorithm may be used to train the retention prediction model based on the second target user data.

[0179] The second target user data can be processed using a big data computing framework (mapreduce+hive). Specifically, hive can be used to collect the second target user data, the model can be encapsulated through mapreduce, and then hive can be used to verify or test the results predicted by the model.

[0180] In an optional implementation, the prediction module 204 trains a retention prediction model based on the second target user data, including:

[0181] Acquire third initial user data of a third target user group, and obtain third target user data according to the third initial user data;

[0182] Using the second target user data and the third target user data as input data of the model, and using the retained labels corresponding to the second target user data and the third target user data as training targets of the model;

[0183] According to the input data of the model and the training target of the model, training is performed based on the Gradient Boosting Decision Tree (GBDT) algorithm to obtain the retention prediction model.

[0184] The third target user group refers to historical users who have given up on renewal. The third initial user data is the same as the first initial user data and the second initial user data, and may also include the time interval of user activities, the time difference between the time of the user's last activity and the current time, the user's registration time, etc. The third registration time of each of the third target users is obtained, and the third time impact factor of the corresponding third initial user data is calculated according to the third registration time, and the third time impact factor is used to update the third initial user data to the third target user data.

[0185] The retention label of the second target user data is recorded as "1", and the retention label of the third target user data is recorded as "0". The second target user data and the third target user data are used as training data sets to train the gradient boosting decision tree to obtain a retention prediction model, and then the retention prediction model is used to perform a binary classification prediction on the first target user data. That is, it is predicted whether the retention label of the first target user data is "1" or "0".

[0186] In the above optional implementation, the retention label is continuously fitted to the target for training. At the same time, the GBDT algorithm focuses on training samples with large residuals in each iteration of the iteration process, and then obtains multiple tree models. The final result is obtained by fusing multiple models, which fully considers the influence of residuals and can make the training effect of the retention prediction model better. When the retention prediction model with better training effect is used to perform retention prediction on the first target user data, a more accurate retention prediction label can be obtained.

[0187] The recommendation module 205 is used to recommend products to the first target user group based on the retention prediction tag and the similarity.

[0188] Users' consumption behavior is an important channel for companies to make profits. The larger the user base, the greater the profit space for the company. When the rate of user loss is lower than the rate of acquiring new users, the number of users is increasing. Predicting possible retained users in advance and making product recommendations to retained users in a timely manner are of great practical significance for reducing user churn rate.

[0189] In an optional implementation, the recommendation module 205 recommends products to the first target user group according to the retention prediction tag and the similarity, including:

[0190] Obtaining a target retention prediction label from the retention prediction labels;

[0191] Acquire a first target user corresponding to the target retention prediction label in the first target user group as a user to be recommended;

[0192] Sorting the similarities corresponding to the users to be recommended;

[0193] Determine the second target users corresponding to the top N similarities as matching users of the user to be recommended;

[0194] Recommend products to the user to be recommended based on the products purchased by the matching user.

[0195] Among them, the target retention prediction label is "1".

[0196] Since the first target users with the target retention prediction label have a greater probability of being retained and becoming the main force of economic support for the enterprise, the factor only makes personalized recommendations of products to the first target users with the target retention prediction label in the first target user group.

[0197] The greater the similarity between the first target user group and the second target user group, the more similar their potential interests and hobbies are. Therefore, generating a product list based on the products purchased by the second target user group and recommending it to the first target user group can better meet the purchasing needs of the first target user group and achieve higher recommendation accuracy.

[0198] In the above optional implementation, by combining the retention prediction label and the similarity factor, product recommendations are made for the first target user group, and the effect of product recommendation is better.

[0199] In the embodiment of the present application, after obtaining the first initial user data of the first target user group and the second initial user data of the second target user group, since the personal interests of the users will change continuously with the passage of time, the first initial user data is updated to the first target user data and the second initial user data is updated to the second target user data by calculating the time influence factor, thereby correcting the problem that the target user data cannot be updated in real time due to the time factor, and further by calculating the interest preference factors of the first target user data and the second target user data, the potential interests and hobbies between the target users corresponding to the two target user data are effectively characterized, and the similarity of the corresponding two target user data is calculated based on the interest preference factor, and the accuracy of the similarity is high. The retention prediction model trained based on the second target user data is used to perform retention prediction on the first target user data to obtain a retention prediction label, and finally, by combining the retention prediction label and the similarity factors, product recommendations are performed for the first target user group, and the effect of product recommendation is better.

[0200] Embodiment 3

[0201] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned embodiment of the commodity recommendation method based on artificial intelligence are implemented, such as Figure 1 S11-S15 shown:

[0202] S11, obtaining first initial user data of a first target user group, and obtaining second initial user data of a second target user group;

[0203] S12, calculating a time impact factor and using the time impact factor to update the first initial user data into first target user data and to update the second initial user data into second target user data;

[0204] S13, calculating the interest preference factors of the first target user data and the second target user data, and calculating the similarity between the corresponding first target user data and the second target user data based on the interest preference factors;

[0205] S14: training a retention prediction model based on the second target user data, and using the retention prediction model to perform retention prediction on the first target user data to obtain a retention prediction label;

[0206] S15: Recommend products to the first target user group based on the retention prediction label and the similarity.

[0207] Alternatively, when the computer program is executed by a processor, the functions of each module / unit in the above-mentioned device embodiment are realized, for example Figure 2 Modules 201-205:

[0208] The acquisition module 201 is used to acquire first initial user data of a first target user group and to acquire second initial user data of a second target user group;

[0209] The updating module 202 is used to calculate the time impact factor and use the time impact factor to update the first initial user data to the first target user data and to update the second initial user data to the second target user data;

[0210] The calculation module 203 is used to calculate the interest preference factors of the first target user data and the second target user data, and calculate the similarity between the first target user data and the second target user data based on the interest preference factors;

[0211] The prediction module 204 is used to train a retention prediction model based on the second target user data, and use the retention prediction model to perform retention prediction on the first target user data to obtain a retention prediction label;

[0212] The recommendation module 205 is used to recommend products to the first target user group based on the retention prediction tag and the similarity.

[0213] Embodiment 4

[0214] See also Figure 3 FIG. 2 is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. In a preferred embodiment of the present invention, the electronic device 3 includes a memory 31 , at least one processor 32 , at least one communication bus 33 and a transceiver 34 .

[0215] Those skilled in the art should understand that Figure 3 The structure of the electronic device shown does not constitute a limitation of the embodiments of the present invention, and may be a bus structure or a star structure. The electronic device 3 may also include more or less other hardware or software than shown in the figure, or a different component arrangement.

[0216] In some embodiments, the electronic device 3 is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application-specific integrated circuits, programmable gate arrays, digital processors, and embedded devices. The electronic device 3 may also include client devices, which include but are not limited to any electronic product that can interact with a client through a keyboard, mouse, remote control, touchpad, or voice-controlled device, such as a personal computer, tablet computer, smart phone, digital camera, etc.

[0217] It should be noted that the electronic device 3 is only an example, and other existing or future electronic products that are suitable for the present invention should also be included in the protection scope of the present invention and included here by reference.

[0218] In some embodiments, the memory 31 stores a computer program, and when the computer program is executed by the at least one processor 32, all or part of the steps in the product recommendation method based on artificial intelligence are implemented. The memory 31 includes a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0219] Furthermore, the computer-readable storage medium may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function, etc.; the data storage area may store data created according to the use of the blockchain node, etc.

[0220] The blockchain referred to in this invention is a new application mode of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0221] In some embodiments, the at least one processor 32 is the control core (ControlUnit) of the electronic device 3, and uses various interfaces and lines to connect the various components of the entire electronic device 3, and executes various functions and processes data of the electronic device 3 by running or executing the program or module stored in the memory 31, and calling the data stored in the memory 31. For example, when the at least one processor 32 executes the computer program stored in the memory, it implements all or part of the steps of the commodity recommendation method based on artificial intelligence described in the embodiment of the present invention; or implements all or part of the functions of the commodity recommendation device based on artificial intelligence. The at least one processor 32 can be composed of an integrated circuit, for example, it can be composed of a single packaged integrated circuit, or it can be composed of multiple integrated circuits with the same function or different functions, including one or more central processing units (CPU), microprocessors, digital processing chips, graphics processors, and various control chips.

[0222] In some embodiments, the at least one communication bus 33 is configured to implement connection and communication between the memory 31 and the at least one processor 32, etc.

[0223] Although not shown, the electronic device 3 may also include a power source (such as a battery) for supplying power to each component. Preferably, the power source may be logically connected to the at least one processor 32 through a power management device, so that the power management device can manage charging, discharging, and power consumption management. The power source may also include one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device 3 may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0224] The above-mentioned integrated unit implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, electronic device, or network device, etc.) or a processor to execute a part of the method described in each embodiment of the present invention.

[0225] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0226] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0227] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0228] It is obvious to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any figure mark in the claims should not be regarded as limiting the claims involved. In addition, it is obvious that the word "including" does not exclude other units or, and the singular does not exclude the plural. Multiple units or devices stated in the specification may also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

[0229] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A commodity recommendation method based on artificial intelligence, characterized in that: The method comprises: Acquire first initial user data of a first target user group, and acquire second initial user data of a second target user group; Calculating the time impact factor and using the time impact factor to update the first initial user data to the first target user data and to update the second initial user data to the second target user data, including: obtaining a first registration time of the first target user and obtaining a second registration time of the second target user; calculating a first time impact factor of the first initial user data corresponding to the first registration time, and calculating a second time impact factor of the second initial user data corresponding to the second registration time; using the first time impact factor to update the first initial user data to the first target user data, and using the second time impact factor to update the second initial user data to the second target user data; Calculating the interest preference factor of the first target user data and the second target user data, and calculating the similarity of the corresponding first target user data and the second target user data based on the interest preference factor, including: calculating a first number of identical target user data in the first target user data and the second target user data; calculating a second number of total target user data of the first target user data and the second target user data; calculating the interest preference factor based on the first number and the second number; calculating the Pearson correlation coefficient of the first target user data and the second target user data; and using the interest preference factor to correct the corresponding Pearson correlation coefficient to obtain the similarity; A retention prediction model is trained based on the second target user data, and the retention prediction model is used to perform retention prediction on the first target user data to obtain a retention prediction label, wherein the training of the retention prediction model based on the second target user data includes: obtaining third initial user data of a third target user group, and obtaining third target user data based on the third initial user data; using the second target user data and the third target user data as input data of the model, and using the retention labels corresponding to the second target user data and the third target user data as training targets of the model; training based on a gradient boosting decision tree algorithm according to the input data of the model and the training target of the model to obtain the retention prediction model; Recommend products to the first target user group based on the retention prediction label and the similarity.

2. The commodity recommendation method based on artificial intelligence according to claim 1, characterized in that: The calculating the Pearson correlation coefficient between the first target user data and the second target user data includes: Calculating first mean data according to the first target user data, and calculating second mean data according to the second target user data; Calculating first difference data according to the first target user data and the first mean data, and calculating second difference data according to the second target user data and the second mean data; The Pearson correlation coefficient between the first target user data and the second target user data is calculated based on the first difference data and the second difference data.

3. The commodity recommendation method based on artificial intelligence according to any one of claims 1 to 2, characterized in that: The step of obtaining the first initial user data of the first target user group and obtaining the second initial user data of the second target user group includes: In response to the received product recommendation instruction, parsing the method body of the product recommendation instruction to obtain information carried by the product recommendation instruction; Acquire a preset first tag of the first target user group and a preset second tag of the second target user group from the information; Constructing a first regular expression according to the preset first tag and constructing a second regular expression according to the preset second tag; Determine a first database identifier by using the first regular expression, and determine data traversed from the first database corresponding to the first database identifier as first initial user data; The second regular expression is used to determine a second database identifier, and the data traversed from the second database corresponding to the second database identifier is determined as second initial user data.

4. The commodity recommendation method based on artificial intelligence according to any one of claims 1 to 2, characterized in that: The recommending products for the first target user group according to the retention prediction label and the similarity includes: Obtaining a target retention prediction label from the retention prediction labels; Acquire a first target user corresponding to the target retention prediction label in the first target user group as a user to be recommended; Sorting the similarities corresponding to the users to be recommended; Determine the second target users corresponding to the top N similarities as matching users of the user to be recommended; Recommend products to the user to be recommended based on the products purchased by the matching user.

5. A commodity recommendation device based on artificial intelligence, characterized in that: The device comprises: An acquisition module, used to acquire first initial user data of a first target user group, and to acquire second initial user data of a second target user group; an updating module, configured to calculate a time impact factor and use the time impact factor to update the first initial user data to the first target user data and to update the second initial user data to the second target user data, comprising: obtaining a first registration time of the first target user and obtaining a second registration time of the second target user; calculating a first time impact factor of the first initial user data corresponding to the first registration time, and calculating a second time impact factor of the second initial user data corresponding to the second registration time; using the first time impact factor to update the first initial user data to the first target user data, and using the second time impact factor to update the second initial user data to the second target user data; a calculation module, configured to calculate an interest preference factor of the first target user data and the second target user data, and calculate the similarity of the corresponding first target user data and the second target user data based on the interest preference factor, comprising: calculating a first number of identical target user data in the first target user data and the second target user data; calculating a second number of total target user data of the first target user data and the second target user data; calculating the interest preference factor based on the first number and the second number; calculating the Pearson correlation coefficient of the first target user data and the second target user data; and using the interest preference factor to correct the corresponding Pearson correlation coefficient to obtain the similarity; A prediction module, used for training a retention prediction model based on the second target user data, and using the retention prediction model to perform retention prediction on the first target user data to obtain a retention prediction label, wherein the training of the retention prediction model based on the second target user data includes: obtaining third initial user data of a third target user group, and obtaining third target user data based on the third initial user data; using the second target user data and the third target user data as input data of the model, and using the retention labels corresponding to the second target user data and the third target user data as training targets of the model; training based on a gradient boosting decision tree algorithm according to the input data of the model and the training target of the model to obtain the retention prediction model; A recommendation module is used to recommend products to the first target user group based on the retention prediction label and the similarity.

6. An electronic device, characterized in that: The electronic device includes a processor and a memory, and the processor is used to implement the artificial intelligence-based commodity recommendation method as described in any one of claims 1 to 4 when executing the computer program stored in the memory.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the artificial intelligence-based commodity recommendation method as described in any one of claims 1 to 4 is implemented.